Home Real Estate Zillow Plays the Race Card in Advance of Hosting HUD Secretary Donovan

Zillow Plays the Race Card in Advance of Hosting HUD Secretary Donovan



Last Thursday, Zillow released a report on minority access to mortgages in partnership with the National Urban League. In case you were wondering what the conclusion of the report might have been, it was titled, “A House Divided: How Race Colors the Path to Homeownership“. The topline conclusion comes from Stan Humphries, Zillow’s Chief Economist and its lead point person on policy matters (along with Katie Curnutte, Zillow’s extremely able head of PR):

It’s been more than 50 years since Dr. King fought for equality, yet it is apparent that the American dream of homeownership is not equally shared by all, even today. Our research shows that minority home buyers are encountering difficulties that often aren’t shared by white home buyers, and that even after they achieve the dream, they have been less likely to see a similar return on their investment.

Sounds dire. Horrible. How can this be in 2014?!?

And Marc H. Morial, president and CEO of the National Urban League, beats along with the drumbeat, saying there is “hard work that remains to ensure that all Americans achieve the dream of homeownership and begin to build real and lasting wealth.”

The House Divided report was released to coincide with the live web-streamed Town Hall meeting with HUD Secretary Shaun Donovan, scheduled for tomorrow. The topic? If you guessed how hard minorities have it when it comes to housing, you get a cigar. The event is actually entitled, “Building Equality in Housing”.

There’s just one small problem. These conclusions and these highly charged assertions are simply not supported by the data.

I received the advance copy of the Report last week, and immediately saw some issues. I haven’t written about the Report because I wanted to clarify a few things to my satisfaction. To Zillow’s credit, it has been nothing but responsive and helpful, and I’ve had a chance to speak with Stan Humphries, Katie Curnutte, and with one of the economists on Zillow’s staff who worked on the report, Skylar Olsen. They were extremely helpful despite the fact that it was obvious from the get-go that I was highly skeptical.

In fact, they sent me a subset of the Home Mortgage Disclosure Act (HMDA) dataset they used for the report, and allowed me to do some investigating of my own. (If they sent me the full dataset, I’d have needed a new computer….) Let me cover the Zillow report first, then get into why I think this is a completely unwarranted leap based on extremely thin, if not nonexistent, evidence.

Zillow’s Report

The link above should allow you to download both the Executive Summary and the Full Report.

As I scanned the report, it became quickly obvious that I was to reach the conclusion of Stan’s quote above — that blacks and Hispanics don’t have equal access to mortgages, and the recent housing market collapse hurt them worse.

But things just kept jumping out at me. For example, this chart (A House Divided, p.35):

Screen Shot 2014-01-21 at 10.31.25 AMWhat the chart says is that for conventional mortgages, the median income for a successful white buyer was $84,000, vs. $68,000 and $62,000 for black and Hispanic buyers, respectively. For applications that were denied by banks, the median income was $62,000, $48,000, and $48,000 respectively for whites, blacks, and Hispanics.

But look at the Asian numbers. The median income of the successful Asian applicant was $96,000 and the unsuccessful Asian applicant was $75,000. In other words, the median income of the Asian denied funding by the banks was higher than the median income of blacks and Hispanics who were granted funding. Plus, a $13,000 gap exists between whites and Asians who get denied funding. In fact, the $75,000 median income of the denied Asian is closer to the $84,000 median income of the successful white applicant than it is to the $62,000 of the white who was denied.

Do banks really hate the AZN that much?

The FHA data was much the same (Id.):

Screen Shot 2014-01-21 at 10.36.50 AMFor an Asian to be successful at getting an FHA loan, he had to have $5,000 more than a white applicant, $12,000 more in income than a black applicant, and $15,000 more than a Hispanic applicant.

We can draw the same conclusions for whites. The white applicant for a conventional mortgage with $62,000 in income was denied, while the Hispanic applicant for a conventional mortgage was accepted. If we’re going to talk about lack of equality, seems to me that the conclusion needs to be the exact opposite of the one drawn in the paper.

Across the board, what the data was suggesting to me was that if there’s evidence of anything, it’s lack of equality and access to whites and Asians, a reverse discrimination by the banks.

Naturally, I asked Zillow about this.

What the Data Actually Says

Turns out, the data available to Zillow is simply inadequate to reach the conclusions drawn, whether by me re: reverse discrimination, or by them re: minority access to mortgages.

For example, HMDA data doesn’t include credit scores. Reasons for denying the loan are given, but one simply cannot say that there’s any disparity in treatment by race.

More importantly, Zillow did not normalize for economic factors at all. What I learned from conversing with the Zillow team is that black and Hispanic median income numbers are lower because they buy in lower-value areas. This map (p.25) is important:

Screen Shot 2014-01-21 at 10.46.28 AMWhat I learned is that because blacks and Hispanics are concentrated in lower-income areas of the country, where home values are lower, they naturally apply for smaller mortgages with lower median incomes that can support those smaller mortgages.

In contrast, when I asked about the Asian data, Stan suggested that the disparity came from the fact that Asians live in high-rent, high-cost urban areas of the country, like New York City, San Francisco, Southern California, and Hawaii, and that therefore, they needed to get more expensive mortgages, which naturally leads to needing far higher median income. In fact, Stan pointed to the fact that Asians have very high Loan Value-to-Income (LTI) Ratio at 2.75.

Of course, Hispanics have an even higher LTI Ratio at 2.78, but the data supposedly suggests that the problem for Hispanics is lower income, while the problem for Asians is high Loan Value, due to where they live, where they want to buy, and hence, the amount of the mortgage.

This is… dubious at best. The Report draws conclusions like this:

Conventional loan applications from black applicants are 2.4 times as likely to be denied compared to an application from a white applicant. A full quarter of all conventional applications from black applicants are rejected by a financial institution, compared to only 10.6 percent of applications from white applicants. Applications from Hispanic applicants don’t fare much better; with a denial rate of 21 percent, Hispanic applications are almost twice as likely to be denied. The denial rate for applications from Asian applicants (13.2 percent) is slightly higher than white applicants’ denial rate. It is not possible to distinguish discrimination in approval procedures from the characteristics of the lender from HMDA data alone. While HMDA data report income and the value of the loan, credit scores and other necessary data are unavailable. (Emphasis added.)

When the data has not bee normalized for median income and loan value, isn’t the above a foregone conclusion? In fact, I added the emphasis because Stan and Skylar both stressed that they are not claiming any sort of invidious discrimination by mortgage brokers, banks, or the FHA. The data simply doesn’t show any discrimination; it merely shows “disparate impact” by race.

But at the same time, given that most of the black and Hispanic applicants were drawn from lower-income areas of the country, and then compared to say Asians who are in high-income areas of the country, the above paragraph should really read like this:

Conventional loan applications from lower-income applicants are 2.4 times as likely to be denied compared to an application from a high-income applicant. A full quarter of all conventional applications from lower-income applicants are rejected by a financial institution, compared to only 10.6 percent of applications from high-income applicants. (Emphasis added.)

Read that way, it is plain that “discrimination” on this basis is not only okay, but prudent. What bank is more likely to lend to the poor versus the rich? In fact, wasn’t the proximate cause of the recent collapse in housing and the resulting Great Recession due to banks not doing enough of this kind of “discrimination” and giving out loans to zero-doc applicants?

Removing Economic Factors

What Zillow needed to have done, if it wanted to reach the kind of strong conclusions that it does, is to try to normalize the data for economic factors as much as possible. Try to isolate race/ethnicity as the sole difference. Which is extremely difficult given the HMDA dataset and what’s available out there.

But not, I think, impossible to do better.

For example, here’s a chart on p.23 of the report:

Screen Shot 2014-01-21 at 11.04.13 AM

Zillow’s conclusion is that over 40% of blacks put down less than 5% as downpayment, reflecting their lower income and lower financial resources as compared to other groups. That’s true. But it’s also obvious that some 25% of blacks put down 20% or more as well. It isn’t as if every single African-American in the US is an impoverished dirt farmer. I mean, Oprah is black, last I looked.

What I wanted to do was to look only at a group of people at one end of the spectrum or the other. I elected to concentrate on the ultra high end, reasoning that while higher-income individuals might “slum it” in lower income areas for a variety of reasons, or make purchases in lower-income areas for investment purposes, the reverse is not true. A janitor of any race isn’t putting in mortgage applications to buy $2M homes in Beverly Hills.

So Zillow sent me a subset of their HMDA dataset at my request (have I mentioned how nice they are?) isolated to some Superzips. I wrote about Superzips last year, but not in the race/ethnicity context. But I think Superzips are a great way to try and remove the economic factors as much as possible, because a Superzip is by definition those zip codes where the income and education levels of the residents are in the top 5% of all the zip codes in the United States. I asked for a dataset concentrating on the Top 20 Superzips, which is now in the top <1% of all of the zip codes.

These zip codes — places like Chappaqua, NY where Bill and Hillary Clinton have a home, Atherton, CA, and Potomac, MD — are where America’s elites of elites live. No one looking to buy a house in these zip codes, and therefore applying for a mortgage, is a struggling factory worker. They’re more likely to be the factory owner, whatever their race.

I ran some numbers myself, because… I like stuff like that. I know, it’s a sickness. But here are my topline findings from looking at the HMDA data limited to the top Superzips.

Data from Superzips

First of all, unsurprisingly, when we get into the top Superzips, the population of blacks and Hispanics drop precipitously. The Superzips are not only high-income, high-education, but they’re also lily-white.

Zillow tells us on p. 32 that blacks and Hispanics make up 4.6% and 17.3% of the total population, respectively. And amongst all mortgage applicants, blacks and Hispanics make up 5.2% and 6.1% of them, respectively.

In the Superzips, blacks and Hispanics make up 0.55% and 1.39% of all mortgage applicants.

[NOTE: Out of the 11,755 mortgage applicants in the Superzips, 2,550 did not list their race/ethnicity. It is possible that many of those applicants are black and Hispanic. But we’ll never know.]

Here’s the topline:

As you can see from the data, the disparity fades away once we limit our data to the elite Superzips. To be fair, the sample size in some cases were tiny (e.g., only two black applicants for a FHA loan were denied), but we do get a better sense of access to mortgages by race, especially for conventional loans.

(NOTE: Given the nature of these zip codes, one would expect that FHA loans are not as big a factor, and that was true. Only 101 out of the 11,755 mortgage applicants were for FHA loans at all.)

Looking at the conventional loans, we see that the disparity in median income for originated and denied loans start to fade away. Successful white and black applicants have nearly identical numbers:

  • White: $230,000 median income, 2.21 LTI Ratio
  • Black: $231,000 median income, 2.28 LTI Ratio

Asians and Hispanics apparently go for more house on lower median income, but it doesn’t seem crazy at 2.58 and 2.67 respectively.

What about those applicants who were denied? Again, the numbers are more or less in line with each other:

  • White: $208,000 median income, 2.74 LTI Ratio
  • Black: $181,000 median income, 2.67 LTI Ratio
  • Asian: $169,000 median income, 3.13 LTI Ratio
  • Hispanic: $242,500 median income, 2.92 LTI Ratio

The Asian number jumps out, but… those folks are clearly going for too much home. House-rich, cash-poor is fine, if you can swing it. But it’s hard to blame the bank for refusing to fund that mortgage. It seems rather unlikely that we’re talking about “unequal access” to the dream of homeownership for Asian doctors in Bethesda, MD.

What about likelihood of denial?

  • White: 695 Denied, 5644 Originated, 10.9% denial rate
  • Black: 13 Denied, 25 Originated, 34.2% denial rate
  • Asian: 123 Denied, 1067 Originated, 10.3% denial rate
  • Hispanic: 20 Denied, 107 Originated, 15.7% denial rate

Blacks are more than three times as likely to be denied conventional mortgages than whites or Asians, and more than twice as likely to be denied than Hispanics. So that’s something. But 7 out of the 13 denials of black applicants for conventional mortgages listed “Credit History” as the reason for denial.

Isn’t that exactly why banks should deny someone a $1.5 million loan, no matter the race? Because they have poor credit?

So the best we can do with the above is to suggest that the black upper class, those who have made it and could think about buying a house in Chappaqua or Old Greenwich, may have worse credit than whites or Asians. They may have come further, struggled more when young. Great for them, and certainly, they may be role models we all need, instead of rappers and basketball players… but as evidence that the American dream of homeownership is not shared by all, it’s pretty thin soup.

What, If Anything, Can We Conclude?

So… that was fun for numbers nerds. But what, if anything, could one conclude from all this?

Just because the super-elites of all races appear to face similar hurdles doesn’t mean there isn’t discrimination or invidious racist stuff going on in the mortgage market. Of course, Zillow takes pains to stress that it is not suggesting there is any bias like that; merely that the data suggests disparate impact on blacks and Hispanics.

But the Superzip analysis shows that once you control — somewhat, in a roundabout way, and imperfectly — for economic factors, there really isn’t a major difference between whites, blacks, Asians, and Hispanics. Zillow’s conclusion about A House Divided, and 50 years after Martin Luther King Jr., and so on and so forth turns out to be empty rhetoric, unsupported by the actual data. (Or, to be charitable, supported only on the thinnest stalks of reed possible.)

What Zillow’s data actually suggests is that more blacks and Hispanics live in poorer areas of the country that became bubblicious in the mid-90’s and remain mired in post-bubble recession: Florida, Nevada, Arizona, California.

What Zillow’s data actually suggests is that banks don’t want to make mortgages to people with bad credit or who are going to struggle to make their house payments (LTI Ratio). What the data shows is that the banks are actually doing exactly as everyone from the Left to the Right have said they should do: make more prudent loans.

You know, like this guy says:

The disparate impact, if there is one, appears to be on the lower-income families of all races. Now that is a conversation we might want to have, related to whether the banks have pulled back too much, whether their lending standards are too tight, etc. etc. That would be a useful conversation.

But because Zillow and The National Urban League played the Race Card, it seems unlikely that we are going to have that conversation at all.

Tomorrow’s Town Hall with Secretary Donovan will be colored by the Report and its thinly/not-at-all supported conclusions about minority access to homeownership. Politicians are bound to jump on this issue and start QQing about how the evil bankers are discriminating against the black man. Or the Latino man. Depending on which constituent said politician is trying to win over, I suppose.

I appreciate the efforts by the Zillow team. I further appreciate their professionalism and willingness to debate, and even to provide datasets to a guy who clearly disagrees with their conclusion.

I also note that this Town Hall with Secretary Donovan marks yet another major win for Zillow in the public policy arena. I can only assume that the folks at NAR are gnashing their teeth once more and trying to figure out how they, and their official portal Realtor.com, can reclaim that ground.

So this paper, A House Divided, and the Town Hall will be significant markers in the public policy arena on housing. Which is why I find it so lamentable that Zillow overextended itself in drawing conclusions, which may satisfy The Urban League and others who have a vested interest in finding racism and discrimination everywhere but are simply unsupported by the data, rather than looking at the data closer to make sure of what it says.

After all, if a solo non-economist blogger could do what I did, I know the world-class data team at Zillow could do even better.

Zillow and the Urban League played the race card, and too early and without enough support. That will taint the conversation. Too bad for us all. I expect better next time around.